Advanced methods of signal processing for Power Quality assessment.

Abstract

The aim of this work is by using artificial neural networks (ANNs) compare six regression algorithms supported by 14 power-quality features, based on higher-order statistics (HOS). In addition, we have combined time and frequency domain estimators to deal with non-stationary measurement sequences; the final target is to implement the system in a smart grid to guarantee compatibility between all the equipment connected. The main results were based on spectral kurtosis measurements, which easily adapt to the impulsive nature of the power quality events. Through these results we have verified that the developed technique is capable of offering interesting results at classifying power quality (PQ) disturbance. We can conclude that using radial basis networks, generalized regression and multilayer perceptron, we have obtained the best results mainly due to the non-linear nature of data.